ESTIMATING EFFECTIVE PERFORMANCE GRADE OF ASPHALT BINDERS IN HIGH RAP MIXTURES USING DIFFERENT METHODOLOGIES: CASE STUDY (Paper No.

Transcription

1 E. Hajj, L. Loria, P. Sebaaly ESTIMATING EFFECTIVE PERFORMANCE GRADE OF ASPHALT BINDERS IN HIGH RAP MIXTURES USING DIFFERENT METHODOLOGIES: CASE STUDY (Paper No. -) By Elie Y. Hajj, Ph.D. Assistant Professor Western Regional Superpave Center Department of Civil & Environmental Engineering University of Nevada Reno N. Virginia St./ MS Reno, Nevada, Phone: --0 Fax: -- elieh@unr.edu Luis Guillermo Loría Salazar, Ph.D. General Coordinator Transportation Infrastructure Program LanammeUCR Phone: (0) - Fax: (0) - LUIS.LORIASALAZAR@ucr.ac.cr Peter E. Sebaaly, Ph.D., P.E. Director/Professor Western Regional Superpave Center Department of Civil & Environmental Engineering University of Nevada Reno N. Virginia St./ MS Reno, Nevada, Phone: -- Fax: -- psebaaly@unr.edu Word Count: 0 + *0 + *0 = 0 Submission Date: August, 0 Revision Date: November, 0

2 E. Hajj, L. Loria, P. Sebaaly 0 0 ABSTRACT In 00 HMA pavement sections containing 0, and 0% RAP were built in a collaborative effort between the Manitoba Infrastructure and Transportation and the Asphalt Research Consortium. There were two types of 0% RAP mixtures evaluated: one with no grade change in asphalt binder (PG-) from the mixtures with lower RAP contents and a second mixture with a grade change in asphalt binder (PG-). The effective binder properties of the evaluated field-produced mixtures were determined by various methodologies including; grading of the recovered binders, blending chart process, mortar procedure, and backcalculation of binder properties from measured dynamic modulus of mixtures using the Hirsch model and the modified Huet-Sayegh model. Overall, good correlations were observed between the estimated critical temperatures from the blending chart process and the measured ones from the recovered asphalt binders. Among the various evaluated methods, the mortar procedure provided promising results when used to estimate the mixture binder properties at critical pavement temperatures. The findings from the mortar procedure were consistent with the mixtures resistance to thermal cracking and their current field performance. The procedure indicated that a partial blending is occurring between the virgin and RAP binders of the evaluated mixtures. Though some difficulties arise with the use of Hirsch model, the backcalculated binder shear moduli were reasonable. The modified Huet-Sayegh model requires further evaluation to assess the true relationship between the characteristic times of the binders and mixtures. Keywords: Recycled asphalt pavement, extraction, mortar, BBR, Hirsh model, Huet-Sayegh, blending chart, master curve

3 E. Hajj, L. Loria, P. Sebaaly INTRODUCTION In the United States (U.S.) alone, demand for asphalt products is expected to increase.% annually to nearly 0 million tons, or million barrels, in 0. About % of the total asphalt demand will go toward asphalt paving applications (). Alongside demand, consistently rising construction, shipping, and manufacturing costs have generated a need for serious conservation movements within the asphalt industry. Beginning in the mid-0 s, recycling asphalt became a widely accepted practice, with more than 0 States practicing some form of pavement recycling between and. Today, all 0 States accept the use of recycled pavements in some form, as either aggregate material or as binder replacement (). Consequently, an effective mix design for hot mix asphalt (HMA) containing recycled asphalt pavement (RAP) is deemed necessary. The mix design should account for the presence of the stiff, oxidized reclaimed binder from the RAP and its effect on the virgin binder properties at all critical pavement temperatures, particularly, when dealing with high RAP content mixtures (i.e. more than %). Evaluating the RAP materials consists of measuring the properties of the binder and aggregates of the RAP mixture. Several research studies have been conducted to identify the best methods for separating and testing the binder and aggregates of the RAP materials, but there have not been any standard procedures that agencies can use on a routine basis. In the case of the aggregates in the RAP, the two critical properties are: gradation and specific gravity. The gradation of the aggregates in the RAP materials can be easily evaluated through the extraction process. Determining the specific gravity of the RAP aggregates represents a challenge. Several techniques have been used in the past, but there is not an established standard procedure to evaluate RAP aggregate properties. In the case of the binder in the RAP, the two critical properties are: binder content and binder properties. The binder content of the RAP can be easily identified through the extraction process. However, measuring the properties of the binder is still a challenging process. Methods currently available for quantifying the effect of the reclaimed binder on the mixture binder properties include, but are not limited to, chemical extraction and recovery of the reclaimed binder, backcalculation from testing gyratory compacted mixture samples, and testing samples cut from gyratory compacted samples in the Bending Beam Rheometer (BBR). Additionally, research conducted within the Asphalt Research Consortium (ARC) (,, ) has resulted in a testing procedure in which mortar and binder samples are tested in the BBR and Dynamic Shear Rheometer (DSR) to quantify the effect of RAP binder on the fresh (i.e. virgin) binder continuous grade profile, allowing for an estimation of mixture binder properties at critical pavement temperatures. This approach has been developed with the intent to minimize labor and eliminate the need for binder extraction. Though it is currently the most widely used method, extracting and recovering the asphalt binder from the RAP materials creates the fundamental issues of the impact of the extraction/recovery process on the properties of the recovered binder and the health/environmental impact of the chemicals used in the process. The backcalculation process was introduced in 00 by Christensen et al. () after refining the Hirsch model to predict the dynamic modulus of HMA using the shear modulus of the binder used and volumetrics of the mix. Unlike the Witczak equation for dynamic modulus (), which is based entirely on a regression approach, the Hirsch model is based on the theory of composite materials.

4 E. Hajj, L. Loria, P. Sebaaly In 00, Daniel and Mogawer () assessed various methods to determine the binder grade in mixtures designed with RAP from the properties of the mixture itself. The study evaluated mixtures with 0, 0, and 0% RAP with a virgin PG- binder. Furthermore, virgin mixtures with PG-, PG0- and PG- binders were evaluated. Testing included dynamic modulus, creep compliance and strength tests in the indirect tensile mode. Partial shear modulus master curves were measured on the recovered binder from each mixture. Several methods of estimating the effective PG grade of the binder in the RAP-containing mixture were examined: linear blending chart, empirical method using measured mixture properties, and backcalculation using the Hirsch model. The researchers found that the empirically based methods of interpolating values from measured mixture properties require an extensive amount of testing in the laboratory. Additionally, the relationship between material properties and PG grade must be established for each type of mixture (gradation, asphalt content). The most promising method for determining the effective PG grade of a mixture was using the Hirsch model to backcalculate the binder stiffness from the measured mixture modulus. Zoftka et al. () investigated the use of the BBR test on thin beams of asphalt mixture to estimate the asphalt binder properties in RAP-containing mixtures. It was shown that the Hirsch model can be used to backcalculate the asphalt binder stiffness at low temperature from the creep compliance (and stiffness) of HMA mixtures (thin beams). Additional research was needed to further investigate the Hirsch model and refine it to obtain reasonable stiffness values and binder m-values. In an attempt to encourage the construction of more sustainable asphalt pavements, the Manitoba Infrastructure and Transportation (MIT) built in collaboration with the Asphalt Research Consortium (ARC), in 00, pavement sections with high RAP contents on Provincial Trunk Highway in Manitoba, Canada. The pavement sections are intended to provide necessary information for assessing the feasibility of using HMA mixtures with high RAP content as surface layers in cold weather regions such as Manitoba, Canada. This paper focused on the effect of reclaimed binder on the field-produced mixture binder properties from Manitoba. OBJECTIVE The overall objective of this paper was to evaluate the binder properties of field-produced mixtures from Manitoba, Canada using the following methodologies: Grading of asphalt binders recovered from the field-produced HMA mixtures with RAP. Estimating the blended asphalt binder grade using the blending chart process described in the AASHTO M (0) testing procedure. Estimating the mixture binder properties using the mortar testing procedure developed under the ARC research. Backcalculating the mixture binder properties from measured mixture dynamic modulus. Backcalculating the mixture binder properties using the modified Huet-Sayegh model. Additionally, the low performance temperatures of the field-produced mixture binders were compared to the fracture temperatures of the corresponding mixtures.

5 E. Hajj, L. Loria, P. Sebaaly DESCRIPTION OF MANITOBA PROJECT The project is located on Provincial Highway between Gimli and Hnausa in Manitoba, Canada. The total project length is approximately miles, with the comparative pavement site accounting for miles of the project. The evaluated sections were constructed in September 00 and each consisted of two -inch lifts with: conventional HMA (i.e. 0% RAP), HMA with % RAP, or HMA with 0% RAP. Two different sections containing 0% RAP were made: one with no grade change for the new asphalt and one with a grade change. The pavement sections were laid on top of a -inch HMA with 0% RAP that was constructed the year before (i.e. 00), which is on top of a base and a subgrade. All four evaluated mixtures in the top two lifts consisted of a dense-graded asphalt mixture manufactured with a Pen 0-00 asphalt binder (average penetration of and graded as PG-) except for the 0% RAP mixture with grade change that was manufactured with a Pen asphalt binder (average penetration of and graded as PG-). The target binder grade for the project location was a Pen The average in-place densities for the HMA with 0,, 0 (Pen 0-00) and 0% (Pen 00-00) RAP were, respectively,,, and %. Loose mixtures were sampled during paving of the top lift from the paving auger at the project site. Those mixtures were referred to as field-produced mixtures and are labeled F-0%- 0, F-%-0, F-0%-0 and F-0%-00. For instance, The F-0%-0 mixture represents the field mixture with 0% RAP manufactured with the Pen 0-00 asphalt binder (i.e. PG-) while the F-0%-00 mixture represents the field mixture with 0% RAP manufactured with the Pen asphalt binder (i.e. PG-). Additionally, virgin asphalt binders and RAP materials (i.e. 00% RAP mixture) were sampled during production at the plant location. DYNAMIC MODULUS TESTING OF ASPHALT MIXTURES The dynamic modulus ( E* ) test was performed according to AASHTO TP-0 (0) for all mixtures including the 00% RAP mixture. Figure a shows the E* master curves for the various mixtures that were determined using the data from three replicates. The E* values at 0 Hz loading frequency and C are presented in Figure b. Examining the E* data leads to the observation that the E* property of field-produced mixtures increases with the increase in RAP content. A reduction in the E* property was observed for the mixture with 0% RAP and PG- when compared to the mixture with 0% RAP and PG-. The highest E* was observed for the 00% RAP mixture closely followed with the mixture with 0% RAP and PG-. The measured dynamic modulus master curves will be used to estimate the effective performance grade of asphalt binders in the various RAP-containing mixtures using the Hirsch model and the modified Huet-Sayegh model theories. SHEAR MODULUS MASTER CURVES OF ASPHALT BINDERS All binders were extracted from the field-produced and RAP mixtures using a centrifuge (AASHTO T (0)) and recovered using a rotary evaporator (ASTM D0 ()) with a solution of % Toluene and % Ethanol by volume (ToE). Then, the shear modulus and phase angle master curves of the various recovered asphalt binders were determined using the

6 E. Hajj, L. Loria, P. Sebaaly Christensen-Anderson-Sharrock (CAS) model shown in Equation. All binder testing was conducted at the Western Research Institute (WRI) laboratory. / / 0/ / where, G b * = complex shear modulus G o = glassy shear modulus = reduced frequency (rad/sec) o = crossover frequency (rad/sec) = width parameter δ = phase angle constant The temperature dependency of the shift factors, a T, is modeled using the Kaelble modification of the Williams-Landel-Ferry (WLF) model. The reduced time,, is then determined by dividing the physical time by the shift factor. log where, a T = shift factor t = physical time = reduced time C, C = model constants T r = reference temperature ( C) T k = inflexion temperature ( C) T = analysis temperature ( C) (a) (b) (c) log () Figure presents the shear modulus master curves of the various recovered binders at 0 C. Examining the master curves data leads to the observation that the G b * property of fieldproduced mixtures increases with the increase in RAP content. A reduction in the G b * property was observed for the mixture with 0% RAP and PG- when compared to the mixture with 0% RAP and PG-. The highest G b * was observed for the 00% RAP mixture. In summary, the same trends were observed for the moduli of the binders and mixtures. GRADING OF VIRGIN AND RECOVERED ASPHALT BINDERS The Superpave Performance Grading (PG) binder system (AASHTO M0 (0)) was used to grade the virgin binders, RAP binder and recovered blended binders from the various loose fieldproduced mixtures. As mentioned before, all recovered binders were extracted and recovered using ToE. The recovered binders were graded by testing them as original, short-term aged through the Rolling Thin Film Oven (RTFO), and long-term aged through the Pressure Aging

7 E. Hajj, L. Loria, P. Sebaaly Vessel (PAV). Figure and Table summarizes the high, intermediate and low critical temperatures of the various binders as well as the Superpave PG grades. Critical temperatures are temperatures at which a binder just meets the appropriate specified Superpave criteria. Based on the presented data the following observations can be made: The low critical temperatures of both virgin binders (i.e. Pen 0-00 and Pen 00-00) were only within C of each other. However, the high critical temperatures were within C of each other. In the case of field-produced mixtures manufactured with PG- asphalt binder, the increase in RAP content in a mixture resulted in warmer high and low critical temperatures for the recovered asphalt binders. The recovered blended asphalt binder from the F-0%-00 mixture was softer than the recovered blended asphalt binder from the F-0%-0 mixture by about C. The recovered binders from the mixtures containing 0 and % RAP met the target grade of PG- for the project location. The recovered binders from the mixtures with 0% RAP met or exceeded the target high PG of C but failed to meet the target low PG of - C. This observation was true for both mixtures with and without grade change. The use of softer asphalt binder (i.e. PG-) with the 0% RAP mixture did not improve the low performance temperature of the blended asphalt binder in the mixture enough to meet the target low PG. As mentioned before, the low critical temperature of both virgin binders were only within C of each other. Due to the limited availability of various asphalt binder types in the project vicinity, the only available commercial soft binder was used in this study (i.e. Pen 00-00). It should be noted that MIT has been using the Pen asphalt binder in their asphalt base layers with mixtures containing up to 0% RAP. ESTIMATION OF ASPHALT BINDER PROPERTIES Blending Chart Process The blending chart process which is described in AASHTO M (0) is used to determine the virgin binder grade or the maximum amount of RAP that can be used. The concept assumes that a 00% blend of the virgin binder and the binder from the RAP occurs. For a known RAP percentage, the critical temperatures of the virgin asphalt binder at high, intermediate, and low properties can be determined using Equation which is based on the assumption that a linear relationship exists between the critical temperature and the RAP content. % % where, T Blend = critical temperature of blended asphalt binder T virgin = critical temperature of virgin asphalt binder T RAP = critical temperature of recovered RAP binder %RAPbinder = percent RAP binder in the RAP expressed as a decimal The measured critical temperatures of the virgin and recovered RAP binders were used to estimate the grade of the blended asphalt binder in the mixtures with different RAP contents. ()

8 E. Hajj, L. Loria, P. Sebaaly Figure and Table show the blending chart results for the critical temperatures along with the PG grades of the blended binders. The estimated critical temperatures from the blending chart process were in the vicinity of the critical temperatures measured for the recovered blended asphalt binders from the various mixtures. Comparing the PG from the blending chart analysis to the PG of the recovered blended binders, the following observations were found: For mixtures with % RAP, the estimated grade for the blended asphalt binder was PG- and consistent with the PG measured from the field-produced mixture (i.e. F-%-0). For mixtures with 0% RAP, the estimated grade for the blended asphalt binder was PG- and was consistent with the PG measured from the field-produced mixture manufactured with PG-. The data show that the blending chart process can sometimes underestimate or overestimate the critical temperatures of the recovered binder by ºC. Mortar Testing Process Research conducted within the Asphalt Research Consortium (ARC) has resulted in a testing procedure in which mortar and binder samples are tested in the Bending Beam Rheometer (BBR) and Dynamic Shear Rheometer (DSR) to quantify the effect of RAP binder on the virgin binder (referred to as fresh binder) continuous grade profile, allowing for an estimation of mixture binder properties at critical pavement temperatures (,, ). The testing procedure allows users to estimate the performance grade of the mixture binder given any amount of RAP binder replacement within the mixture. In the proposed analysis procedure, three samples are tested at low, intermediate and high critical temperatures. These samples include one fresh binder and two mortar samples; the mortars are comprised of the same fresh binder and a single gradation aggregate source. The procedure is founded on the idea that if the two mortar samples are prepared with identical gradation and identical total asphalt content (maintaining the same constituents), but one mortar sample contains a percentage of reclaimed binder (replacing an identical percentage of fresh binder), then any difference in properties between the mortar samples can be directly attributed to the percentage of reclaimed binder. If the properties of the fresh binder (also used for making the mortar samples) are also known, then the change in properties of the fresh binder due to blending with the reclaimed binder can be isolated. That is, if the continuous grade of the fresh binder is known, the mortar testing will allow for the estimation of the resulting continuous grade of the fresh binder-reclaimed binder blend. Figure illustrates the mortar preparation process and the distinction between the above-mentioned mortars for low temperature characterization (). The R00 RAP material in Figure consists of the RAP material passing the #0 sieve and retained on the #00 sieve. The Burned R00 Material consists of extracted RAP aggregate using an ignition oven. In summary, this procedure allows for the direct characterization of the reclaimed binder and eliminates the need for chemical extraction and recovery. In the analysis, backcalculation from mortar to binder is not adopted and all calculations are in terms of binder or blended binder only. Figure and Table summarize the critical temperatures determined using the mortar procedure for the and 0% RAP-containing mixtures. In the case of field-produced mixtures

9 E. Hajj, L. Loria, P. Sebaaly manufactured with PG- asphalt binder, the increase in RAP content in the mixture resulted in warmer high and low critical temperatures for the asphalt binders. The critical temperatures for the asphalt binder from the F-0%-00 mixture were softer than those of the asphalt binder from the F-0%-0 mixture by to C. For all mixtures, the mortar procedure resulted in critical temperatures that are lower than those measured on recovered binders or determined using the blending chart process. The low critical temperatures determined using the mortar procedure for the 0% RAP mixtures were significantly cooler (by to C) than those measured on the recovered binders or estimated using the blending chart process. This observation was further examined through the resistance to thermal cracking of the mixtures using the thermal stress restrained specimen test (TSRST). The results are presented later on in the paper. Hirsch Model Approach The original model was developed by Hirsch in the late 0 s to calculate the modulus of elasticity of Portland cement concrete based on the theory of composite materials. In 00, the model was refined by Christensen et al. () to predict the E* of HMA using the G* binder of the binder used and volumetrics of the mixture. The Hirsch model is shown below:,00,000, / (),,.. () where, E* mixture = dynamic modulus for the asphalt mixture, psi G* binder = shear modulus for the binder, psi P c = contact volume VMA = voids in the mineral aggregate, percent VFA = voids filled with asphalt, percent The phase angle in shear can be estimated from the logarithm of P c using the following equation:.. () Table summarizes the critical temperatures for the asphalt binders determined using the Hirsh model. The critical temperatures were determined at the frequency of 0 radians/sec which corresponds to the frequency used to grade binders. Overall, the Hirsh model resulted in high critical temperatures that are softer than those of the recovered binders. It should be noted that since it is not possible to test the mixtures at the high PG grade temperatures, the measured mixture properties were shifted to the higher temperatures using the time-temperature superposition principle. Additionally, some difficulties were faced with the use of the Hirsch model to backcalculate the shear modulus of the binder at high temperatures,

10 E. Hajj, L. Loria, P. Sebaaly particularly with the relatively soft virgin binders used in this study. Due to limitations inherent in the formulation of the Hirsch model itself, the G* binder at high temperatures were estimated from the backcalculated binder properties at relatively lower temperatures (i.e. from the linear relationship between the logarithm of backcalculated G* binder and temperature). Modified Huet-Sayegh Model Approach Di Benedetto and Olard (, ) proposed the modified Huet-sayegh model as an alternative to determine the HMA response at various test frequencies and temperatures. The advantage of the model is that it applies to both binders and mixtures and it provides both the real and imaginary components of the dynamic modulus. The model is also called SPD ( Springs, Parabolics, Dashpot) and is based on a simple combination of physical elements (spring, dashpot and parabolic element). The Huet-Sayegh model () has been adapted by adding a linear dashpot in series with the two parabolic elements and the spring of rigidity E -E 0 (see Figure a). As regards to binders, for which the minimum asymptotic stiffness (referred to as the static modulus ) E 0 is equal to 0, the model is equivalent to a linear dashpot at very low frequency. The SPD model has a continuous spectrum (can be represented by an infinity of Kelvin-Voïgt elements in series or Maxell elements in parallel) and is presented using the following equation (). E iωτ () where, i = complex number (i = -) E = limit of the complex modulus for ωτ (Glassy modulus), E 0 = limit of the complex modulus for ωτ 0, h, k = exponents such that 0 < k < h <, δ = dimensionless constant, β = dimensionless parameter introduced to take into account the newtonian viscosity of the linear dashpot τ = characteristic time varying with temperature accounting for the Time Temperature Superposition Principle (TTSP), and, ω = π frequency. To entirely determine the linear viscoelastic behavior of the considered material only seven constants (δ, k, h, E, E 0, β and τ 0 ) are needed at a given temperature. In the case of binders, the experimental static modulus is very close to zero; for this reason, E 0 can be taken equal to zero and the number of constants of the model can be reduced to six for binders. Assuming the material follows a linear viscoelastic thermorheologically simple behavior (meaning that the TTSP holds), then only the τ parameter depends on temperature. Di Benedetto et al. () presented the relationship between the binder and the mixture complex moduli (Equation ). Hence, if the mixture complex modulus is known at a given temperature, the binder complex modulus at the same temperature can be determined if E 0binder, E binder and are known. Olard and Di Benedetto (,, ) and Falchetto et al. () reported good correlations between the characteristic times of the mixtures and binders with similar values. In this paper, the relationship between the characteristic times of the mixtures and

11 E. Hajj, L. Loria, P. Sebaaly binders were examined to backcalculate the binder properties from the mixture dynamic modulus., 0, τ T 0 α τ T where, E* mix = complex modulus of the mixture, E* binder = complex modulus of the binder, E mix = glassy modulus of the mixture, E 0mix = static modulus of the mixture, E binder = glassy modulus of the binder, E 0binder = static modulus of the binder, T = temperature, and, α = regression coefficient depending on mixture and aging. The constants required for all five mixtures and recovered binders at a reference temperature of ºC were determined with a minimization process from the experimentally measured dynamic modulus and shear modulus data, respectively. The results for both mixtures and binders are presented in Table. Each mixture showed the same parameters δ, k, h and β of the associated binder while only the static and glassy modulus (E 0 and E ) and τ 0 seemed to be specific to binders and mixtures. The findings are consistent with those reported in other literature (, ). Figure b shows the relationship between the characteristic times of the five binders and those of the corresponding field-produced mixtures for a reference temperature of ºC. A good correlation was observed between the characteristic times at ºC (R of 0.). An overall α- parameter value of. was observed for all evaluated mixtures. Though the α value was expected to be constant with temperature, it increased as the temperature increased. The relationship between α and temperature is presented in Figure c and Equation. = 0.000(T) (T) +. (R = 0.) () Equations and were used to estimate the characteristic times of binders from those of mixtures. Subsequently, the SPD model parameters in Table were used to estimate the high critical temperatures of the binders at 0 radian/seconds. The predicted temperatures are shown in Table. Similar trends were observed for the critical temperatures except for the RAP mixture which showed an unrealistic prediction for the high critical temperature. After careful revision of the data, it was found that the error was mainly due to significant deviations in the - parameter values. COMPARISON BETWEEN THE LOW TEMPERATURE PROPERTIES OF BINDERS AND MIXTURES The resistance of the various mixtures to thermal cracking were measured using the Thermal Stress Restrained Specimen Test (TSRST) (AASHTO TP0-). The test cools down a 0 beam specimen at a rate of 0 C/hour while restraining it from contracting. While (a) (b)

12 E. Hajj, L. Loria, P. Sebaaly the beam is being cooled down, tensile stresses are generated due to the ends being restrained. The HMA mixture would fracture as the internally generated stress exceeds its tensile strength. The temperature at which fracture occurs is referred to as fracture temperature representing the temperature at which the asphalt pavement will develop a transverse crack due to thermal stresses. The resistance of the mixtures to thermal cracking was measured at the long-term aged stage since low-temperature cracking is a long-term pavement distress mode. The aging of the mixtures followed the Superpave recommendation for long-term aging of HMA mixtures, which consisted of subjecting the compacted samples to C for days in a forced draft laboratory oven. All samples were compacted using a kneading compactor to final air-voids of ±0.%. Figure a shows the TSRST fracture temperatures of the evaluated mixtures. All mixtures met the target low temperature performance grade of the asphalt binder (i.e. - C). The fracture temperatures of the mixtures manufactured with the PG- asphalt binder decreased with the increase of RAP content. On the other hand, mixtures manufactured with the PG- asphalt binder exhibited fracture temperatures that are similar to the virgin mixtures (i.e. 0% RAP). Figure b shows the relationship between the TSRST fracture temperatures and the low critical temperatures of the corresponding binder as determined using the various methodologies. The TSRST fracture temperatures of the mixtures with 0% and % RAP were within C of the critical low temperatures of the asphalt binders recovered from the same corresponding mixtures. On the other hand, mixtures containing 0% RAP exhibited fracture temperatures that were colder than the critical low temperatures of the asphalt binders recovered from the same corresponding mixtures by to C. However, very consistent results were observed between the critical temperatures from the mortar procedure and the TSRST fracture temperatures. Therefore, full blending between the virgin and RAP binder does not seem to be evident for the evaluated mixtures. INITIAL FIELD PERFORMANCE In 0, a condition survey of the test sections was performed by the Western Research Institute (WRI) personnel. The inspection revealed no distresses in the 0, and 0% RAP pavements after two years of service. Thermal cracking was not apparent at any of the sections. The lowest air temperatures recorded at the project site during the first two years of service was -. C. Overall, the pavement condition was excellent and uniformly the same along the total length of all test sections. Field performance will continue to be monitored over the next few years, and pavement distress survey and falling weight deflectometer (FWD) data will be collected. CONCLUSIONS Several methods of estimating the effective PG grade of the binder in field-produced mixtures designed with RAP were evaluated in this study. Good correlations were observed between the estimated critical temperatures from the blending chart process and the measured ones from the recovered asphalt binders. In some cases, the blending chart process underestimated or overestimated the critical temperatures of recovered binders by C. Both methods showed significant increase in the binder critical temperatures (i.e. warmer temperatures) with the increase in RAP content, particularly when 0% RAP was used. Specifically, the low critical

13 E. Hajj, L. Loria, P. Sebaaly temperatures of the binders recovered from the 0% RAP mixtures were to C warmer than those recovered from the virgin mixture (i.e. 0% RAP). This observation is mainly related to the assumption that full blending is occurring between the virgin and the RAP binders. The mortar approach showed promising results in determining the effective PG grade of the mixture. Though the method was sensitive to the percent of RAP and the type of virgin binder used in the mixture, it did not result in drastically different mixture binder properties for the 0% and 0% RAP mixtures. Overall, the mortar procedure resulted in high, intermediate and low critical temperatures that are lower (i.e. softer) than those determined for recovered asphalt binders. The mortar results for low critical temperatures were further confirmed with measured fracture temperatures on the various mixtures. The low critical temperatures from the mortar procedure were comparable and consistent with the performance of the mixtures to thermal cracking. Consequently, this procedure may well be indicating that a certain level of blending is occurring between the virgin and RAP binders in a mixture. Hence, further evaluation of the proposed mortar procedure is considered necessary using a larger set of mixtures that cover different binder types and sources at multiple levels of RAP content. The Hirsch model provided reasonable estimates to backcalculate binder shear modulus at high temperature from the measured mixture dynamic modulus. However, some difficulties exist in estimating the high temperature PG grade, particularly when dealing with soft binders like the ones used in this study. The difficulties arise from; first, the difference in temperatures between the dynamic modulus testing and PG grading temperatures; and second, the limitation in the formulation of the Hirsch model at high temperatures and low frequencies. Subsequently, in order to get reasonable estimates, the binder modulus at high temperature was estimated from the backcalculated moduli (using the Hirsch model) at lower temperatures from the mixture dynamic modulus. The modified Huet-Sayegh model was used to estimate the high PG of the asphalt binders from the measured dynamic modulus of the corresponding mixtures. This approach resulted in high critical temperatures that are, respectively, similar and higher than those determined for the recovered binders and using the Hirsch model. The data showed good correlation between the characteristic times of the binders and mixtures, hence, allowing for the determination of the binder properties from the measured dynamic modulus of the mixture. However, this relationship was highly dependent on temperature and need to be verified with other types of mixtures. In summary, this study was specific to the use of RAP in typical mixtures in Manitoba, Canada, and further research is required on different types of RAP mixtures and different virgin PG grades to verify and validate the methods evaluated as part of this study. Particularly, the mortar procedure provided promising results and needs to be further examined. ACKNOWLEDGEMENT The content of this paper is part of the overall effort in the Asphalt Research Consortium (ARC). However, the contents of this report reflect the views of the authors and do not necessarily reflect the official views and policies of the FHWA. The authors gratefully acknowledge support from the FHWA. Additionally, the authors acknowledge the Manitoba Infrastructure and Transportation (MIT) for their involvement and contribution in successfully conducting this study.

14 E. Hajj, L. Loria, P. Sebaaly REFERENCES. Roads and Bridges. U.S. asphalt demand to exceed million tons in 0. Retrieved September 00 from: million-tons-in-0-newspiece, 00.. U.S. Department of Transportation. Pavement Recycling. Retrieved September 00, from FHWA: Swiertz, D., Mahmoud, E, and Bahia, H.U. Estimating the Effect of RAP and RAS on Fresh Binder Low Temperature Properties without Extraction and Recovery. Transportation Research Record No. XXXX, Transportation Research Board, Washington, D.C., 0, pp.xx-yy.. Ma, T., Bahia, H. U., Mahmoud, E., and Hajj, E. Y. Estimating Allowable RAP in Asphalt Mixes to Meet Target Low Temperature PG Requirements. Journal of the Association of Asphalt Paving Technologists, Vol., 00, pp... Swiertz, D.R. Development of a Test Method to Quantify the Effect of RAP and RAS on Blended Binder Properties without Binder Extraction. M.S. Thesis, 00.. Christensen, D.W., Pellinen, T. and Bonaquist, R. F. Hirsch Model for Estimating the Modulus of Asphalt Concrete. Journal of the Association of Asphalt Paving Technologists, Vol., 00, pp. -.. NCHRP. Guide for Mechanistic-Empirical Design of New and Rehabilitated Structures. Final Report for Project -A, National Cooperative Highway Research Program, Transportation Research Board, National Research Council, Washington, D.C., 00.. Daniel, J.S., and Mogawer, P.S. Determining the Effective PG Grade of Binder in RAP Mixes. Final Report No. NETCR, New England Transportation Consortium, C/O Advanced Technology & Manufacturing Center, and University of Massachusetts Dartmouth, 00.. Zofka, A., Marasteanu, M., Clyne, X., and Hoffmand, O. Development of simple asphalt test for determination of RAP blending charts. Final Report No. MN/RC-00-, Minnesota Department of Transportation, American Association of State Highway and Transportation Officials (AASHTO). Standard Specifications for Transportation Materials and Methods of Sampling and Testing. 0 th Edition, Washington, D.C., 00.. American Society of Testing and Materials (ASTM) Designation D 0, Standard Practice for Recovery of Asphalt from Solution Using the Rotary Evaporator, 00.. Olard, F., and Di Benedetto, H. General SPD model and relation between the linear viscoelastic behaviors of bituminous binders and mixes. International Journal of Road Material and Pavement Design, Vol. /, Special Issue, 00, pp. -.. Di Benedetto, H., Olard, F., Sauzéat, C., and Delaporte, B. Linear viscoelastic behaviour of bituminous materials: from binders to mixes. International Journal of Road Material and Pavement Design, Vol., Special Issue, 00, pp. -0. Sayegh, G. Variation des modules de quelques bitumes purs et enrobes bitumineux. Ph.D. Thesis, Faculté des Sciences de l Université de Paris, June.. Olard,F. Comportement thermomécanique des enrobes bitumineux a basses températures relation entre les propriétés du liant et de l enrobe. Ph.D. Thesis, Ecole Nationale des Travaux Publics de l Etat, 00.

15 E. Hajj, L. Loria, P. Sebaaly Falchetto, A. C., Marasteanu, M. O., and Benedetto, H. Analogical Based Approach to Forward and Inverse Problems for Asphalt Materials Characterization at Low Temperatures. To appear in the Journal of the Association of Asphalt Paving Technologists, AAPT, Vol. 0, 0.

16 E. Hajj, L. Loria, P. Sebaaly LIST OF TABLES TABLE Summary of Critical Temperatures for Asphalt Binders TABLE Mixtures and Binders parameters of the SPD Model

17 E. Hajj, L. Loria, P. Sebaaly TABLE Summary of Critical Temperatures for Asphalt Binders Mixture ID RAP Content Virgin Binder True Grade Recovered Binder True Grade Blending Chart Mortar Hirsh Model $ Modified Huet- Sayegh $ High Critical Temperature ( C) F-0%-0 0% F-%-0 % F-0%-0 0% F-0%-00 0% RAP-00% 00% Intermediate Critical Temperature ( C) F-0%-0 0% NA* NA* F-%-0 %.... NA* NA* F-0%-0 0%... NA* NA* F-0%-00 0%.... NA* NA* RAP-00% 00% NA NA* NA* Low Critical Temperature ( C) F-0%-0 0% NA* NA* F-%-0 % NA* NA* F-0%-0 0% NA* NA* F-0%-00 0% NA* NA* RAP-00% 00% NA* NA* * Analysis is not applicable for intermediate and low temperatures, $ Corresponds to G*/sin =. kpa TABLE Mixtures and Binders parameters of the SPD Model Parameters Mixture ID F-0%-0 F-%-0 F-0%-0 F-0%-00 RAP Mixtures Parameters K H E 0 (ksi) E (ksi),0,0,,,,000,000,000,000,000 Log( 0 ( C)) Recovered Binders Parameters k h E 0 (ksi) E (ksi),000,000,000,000,000 Log( 0 ( C)) Parameter at C....0.

18 E. Hajj, L. Loria, P. Sebaaly 0 0 LIST OF FIGURES FIGURE Dynamic modulus at C for field-produced mixtures and RAP mixture (a) master curves (b) at, 0 and Hz. FIGURE Recovered asphalt binders master curves at 0 C. FIGURE Asphalt binders critical temperatures and PG grades. FIGURE Sample preparation procedure for RAP binder characterization. FIGURE (a) Representation of the SPD model, h and k are two parabolic creep elements and dimensionless constant, (b) Comparison between Log( 0-mixture ) and Log( 0-binder ) determined at C and (c) variation of -parameter value with temperature. FIGURE (a) TSRST fracture temperatures of various mixtures and (b) relationship between TSRST fracture temperatures and binders low critical temperatures. (Numbers above bars represent mean values and whiskers represent mean ± % confidence interval)

19 E. Hajj, L. Loria, P. Sebaaly.E+0 Dynamic Modulus E* at C (ksi).e+0.e+0.e+0.e+00.e 0.E 0.E 0.E 0.E 0.E 0.E+00.E+0.E+0.E+0.E+0.E+0.E+0,00 F 0% 0 F % 0 F 0% 0 F 0% 00 RAP 00% (a) Dynamic Modulus E* at C (ksi),00, ,,0 0 0 Reduced frequency (Hz) F 0% 0 F % 0 F 0% 0 F 0% 00 RAP 00% (b) FIGURE Dynamic modulus at C for field-produced mixtures and RAP mixture (a) master curves (b) at, 0 and Hz.

20 E. Hajj, L. Loria, P. Sebaaly Log (G*), psi True PG Grade ( C) Log(Reduced Frequency), Hz F 0% 0 (ToE) F % 0 (ToE) F 0% 0 (ToE) F 0% 00 (ToE) RAP (ToE) FIGURE Recovered asphalt binders master curves at 0 C. PG PG PG 0 PG PG PG PG PG PG PG PG PG PG Pen 0 00 Pen RAP binder F 0% 0 F % 0 F 0% 0 F 0% 00 F % 0 F 0% 0 F 0% 00 F % 0 F 0% 0 F 0% 00 Virgin Binders Recovered Binders Using ToE Blending Chart Process Mortar Testing Process High Temperature Intermediate Temperature Low Temperature FIGURE Asphalt binders critical temperatures and PG grades.

21 E. Hajj, L. Loria, P. Sebaaly FIGURE Sample preparation procedure for RAP binder characterization. 0.0 (a).0 Log ( 0 mixture ) Log ( 0 mixture ) = Log ( 0 binder ) +. R² = Log ( 0 binder ) Parameter Value = 0.000(T) 0.00(T) +. R² = Temperature, T ( C) 0 F 0% 0 F % 0 F 0% 0 (b) (c) FIGURE (a) Representation of the SPD model, h and k are two parabolic creep elements and dimensionless constant, (b) Comparison between Log( 0-mixture ) and Log( 0- binder) determined at C and (c) variation of -parameter value with temperature.

22 E. Hajj, L. Loria, P. Sebaaly TSRST Fracture Temperature, C F 0% 0 F % 0 F 0% 0 F 0% 00 Mixture Type Recovered Binder True Grade Blending Chart Process Mortar Procedure TSRST Fracture Temperature ( C) (a) (b) FIGURE (a) TSRST fracture temperatures of various mixtures and (b) relationship between TSRST fracture temperatures and binders low critical temperatures. (Numbers above bars represent mean values and whiskers represent mean ± % confidence interval) Binder Low Critical Temperature ( C)